Upscaling from leaf to canopy: Improved spectral indices for leaf biochemical traits estimation by minimizing the difference between leaf adaxial and abaxial surfaces

The knowledge of leaf biochemical traits is significant to understand the plant growth and physiological status. Spectral indices have been widely used to assess leaf biochemical traits, while the estimation accuracies at canopy level are frequently lower compared to those at leaf level in part due to the complexity of canopy structures and the variations in optical properties between leaf adaxial and abaxial surfaces. This study thus improved spectral indices with minimizing the effect of the difference between leaf adaxial and abaxial surfaces to assess leaf biochemical traits from leaf to canopy level. The datasets including leaf reflectance and canopy reflectance with corresponding leaf chlorophyll content (Cab), water content (Cw), and dry matter content (Cm) from a wide range of plant species were used. Results showed that there existed a significant difference between leaf abaxial and abaxial reflectance, causing the variation in the relationship between leaf biochemical traits and spectral indices. The published spectral indices exhibited the strong relationships with Cab, Cw, and Cm for either leaf adaxial or abaxial data, while the relationships of spectral indices with Cab, Cw, and Cm from leaf adaxial reflectance were inconsistent with those from leaf abaxial reflectance. The proposed adjusted ratio of difference spectral index (ARDSI) minimized the difference between leaf adaxial and abaxial reflectance for assessing Cab, Cw, and Cm at leaf level with the root mean square error of 9.32 μg cm-2, 0.0050 g cm-2, and 0.0053 g cm-2, respectively. The application of the proposed ARDSI to the canopy level alleviated the effect of the variations of leaf adaxial and abaxial reflectance difference and canopy structures on the estimation of leaf biochemical traits, which thus improved the assessment of Cab, Cw, and Cm at canopy level in the simulated and measured datasets. The proposed approach would advance the applicability of spectral indices to monitor the physiological and functional traits of field crops at different scales.